********************************************************************************************
********************************************************************************************
********************************************************************************************

README FILE FOR REPLICATION FILES FOR R.M. EDGE AND J.B. RUDD,
"REAL-TIME PROPERTIES OF THE FEDERAL RESERVE'S OUTPUT GAP"

July 14, 2015

This file provides documentation for the datasets and programs used to generate the
results shown in Tables 1, 2, and 3 of the paper.

Note that all RATS programs were run using RATS for UNIX, version 7.10; all FAME programs
were run using FAME Release 10.2 (64-bit) on a LINUX system.


********************************************************************************************
********************************************************************************************
********************************************************************************************


DATA FILES AND PROGRAMS FOR GENERATING THE RESULTS IN TABLE 1

DATA FILES: gbgap_all.db (fame format), gbgap_all.rat (rats format), 
            gbgap_all.xls (excel format)
* These three data files are identical to each other; they are just saved in different 
  formats.
* Each file contains 53 series, all but one of the series have names of the form 
  "GB_GAP_GDPXXXX".
* The series that does not have this form is "GB_GAP_RT", which provides the real-time 
  estimates of the Greenbook output gap.
* For all the other series in the file the "XXXX" in series name "GB_GAP_GDPXXXX" is a 
  date; specifically, either the quarter "94Q3" or any quarter between "96Q1" and 08Q3".
  These series are full timeseries of the output gap assumed in the FRB Greenbook. The 
  timing convention (described in the paper) used for our Greenbook gap estimates means 
  that the series with the suffix "94Q3" was the FRB Greenbook output gap estimated for the
  November 1994 Greenbook, the series with the suffix "96Q1" was the FRB Greenbook output 
  gap estimated for the May 1996 Greenbook, and so on.
* The series names "GB_GAP_GDP9413" was provided to us by Athanasios Orphanides as too were
  the observations for the series "GB_GAP_GDPRT" for all quarters prior to 1996:Q1. All of
  the other series in the data files and all of the other observations of "GB_GAP_RT" are
  from Fed datasets.

PROGRAM FILES: table1stats.inp (fame program)
* This program calculates the means, standard deviations, RMSEs, and noise-to-signal ratios
  shown in Table 1.
* The program uses the data file gbgap_all.db for its calculations. (It does not use the
  files in the other formats, which are provided for the convenience of the reader.)

OUTPUT FILES: table1stats_output.txt (text file)
* This file provides the output from table1stats.inp. The numbers generated by 
  table1stats.inp and shown in table1stats_output.txt correspond to those in Table 1.


********************************************************************************************
********************************************************************************************
********************************************************************************************


DATA FILES AND PROGRAMS FOR GENERATING THE PERCENTILE BOOTSTRAP DISTRIBUTIONS FOR THE
TABLE 1 RESULTS

The following data files and programs generate empirical distributions for the
noise-to-signal ratios (NSRs) for the Greenbook gap revisions over the 1980-1992 or
1966-1997 subsamples.  (This is the "first approach" that we use in order to assess the
statistical significance of the observed declines in the NSRs that we find for the
more-recent subsamples--see Section III of the paper.)

DATA FILES: table1_boot_data.rat (RATS format)
* This data file contains three series: GB_GAP_RT (the real-time estimates of the
  Greenbook output gap), GB_GAP_GDP94Q3, and GB_GAP_GDP99Q3.  (These series are the same
  as those described in the preceding section.)

PROGRAM FILES: The following RATS programs use a circular moving block bootstrap to compute
the empirical distributions for the noise-to-signal ratios (NSRs) for the Greenbook gap
revisions over the 1980-1992 or 1966-1997 subsamples.
* pct_boot_66_97_rmse_blk4.prg: Uses RMSE-based NSR over the 1966-97 sample; block size=4
* pct_boot_66_97_rmse_blk5.prg: Uses RMSE-based NSR over the 1966-97 sample; block size=5
* pct_boot_66_97_stddev_blk4.prg: Uses standard-deviation-based NSR over the 1966-97 sample;
                                  block size=4
* pct_boot_66_97_stddev_blk5.prg: Uses standard-deviation-based NSR over the 1966-97 sample;
                                  block size=4
* pct_boot_80_92_rmse.prg:   Uses RMSE-based NSR over the 1980-92 sample; block size=4
* pct_boot_80_92_stddev.prg: Uses standard-deviation-based NSR over the 1980-92 sample;
                             block size=4

OUTPUT FILES: The text files pct_boot_66_97_rmse_blk4.out, pct_boot_66_97_rmse_blk5.out,
etc. contain the output from the corresponding .prg program files (the output takes the
form of a tabulation of the percentiles of the bootstrapped empirical distribution).


********************************************************************************************
********************************************************************************************
********************************************************************************************


DATA FILES AND PROGRAMS FOR GENERATING THE RESULTS IN TABLE 2

There are two steps involved in generating the results in Table 2.  The first step is 
generating the statistical output gap estimates.  The second step is calculating the 
signal-to-noise ratios shown in Table 2 based on the statistical output gap estimates
calculated in the first step.

********************************************************************************************

FIRST STEP IN GENERATING THE RESULTS IN TABLE 2

DATA FILES: gdp6679.db,  gdp8094.db,  gdp9514.db (fame format)
            gdp6679.rat, gdp8094.rat, gdp9514.rat (rats format)
            gdp6679.xls, gdp8094.xls, gdp9514.xls (excel format)
* The three data files gdp6679.db, gdp6679.rat, and gdp6679.xls are identical to each other;
  they are just saved in different formats.  The same is true for the three data files
  gdp8094.db, gdp8094.rat, and gdp8094.xls and the three data files gdp9514.db, 
  gdp9514.rat, and gdp9514.xls.
* The gdp6679.___ files each contain 55 series, all of which have names of the form 
  "GDPXXXX" The "XXXX" in the series name "GDPXXXX" is a date; specifically any quarter
  between "66Q1" and "79Q3".
* The gdp8094.___ files each contain 61 series, all of which have names of the form 
  "GDPXXXX" The "XXXX" in the series name "GDPXXXX" is a date; specifically any quarter
  between "79Q4" and "94Q4".
* The gdp9514.___ files each contain 77 series, all of which have names of the form
  "GDPXXXX" The "XXXX" in the series name "GDPXXXX" is a date; specifically any quarter
  between "95Q1" and "14Q1".
* All of the "GDPXXXX" series in the gdp6679.___, gdp8094.___, and gdp9514.___ files are
  full timeseries of estimates of real GDP as of that point in time. In all but the cases 
  listed in the following bullet point these are exactly the series published by the U.S. 
  Bureau of Economic Analysis (BEA). The source for these series is the St. Louis Fed 
  ALFRED database (see www.alfredstlouisfed.org). The timing convention used for the names 
  of these series is the reference quarter of the data, not the timing of when the data was 
  released.
* The quarters for which the "GDPXXXX" series given in the gdp6679.___, gdp8094.___, and 
  gdp9514.___ files are not exactly those published by the U.S. BEA is when the BEA series
  have some missing data at the start of the time series.  This happens for the following 
  vintages of data 91Q3, 91Q4, 92Q1, and 92Q2 (when the data starts in 1959q1 rather than 
  in 1947q1), 95Q3, 95Q4, 96Q1, 96Q2, 96Q3, and 96Q4 (when the data starts in 1959q1 rather
  than in 1947q1), and 99Q3 and 99Q4 (when the data only starts in 1959q1 rather than 
  1947q1). These vintages are immediately after comprehensive revisions and the missing 
  data reflects the fact that the BEA requires a bit of time before calculating the full 
  revised time series of data.  For these vintages we extrapolate back using the growth
  rates of the GDP series available just prior to the comprehensive revision. This adjust-
  ment is the same as what Orphanides and Van Norden (2002) do with their real-time data.

PROGRAM FILES: hp6614.prg, lqd6614.prg, wat6614.prg, hc6614.prg, bk6614.prg, bn6614.prg
               (all rats programs)
* Each of these program files calculate real-time and final statistical-based output gap 
  measures. The prefixes of the program names indicate which statistical method (or 
  methods) are being used in the program to generate the program's output gaps. The naming 
  convention for these statistical methods are: hp = Hodrick-Prescott filter, wat = Watson 
  Kalman filter model, hc = Harvey and Clark Kalman filter model, bk = Baxter King 
  band-pass filter model, and bn = Beveridge-Nelson trend-cycle decomposition. The program 
  "lqd6614.prg" uses three different de-trending methods; specifically, the linear 
  detrending method, the quadratic detrending method, and the broken trend detrending 
  method.
* The programs use the data files gdp6679.rat, gdp8094.rat, gdp9514.rat for its 
  calculations. (It does not use the files in the other formats, which are provided for 
  the convenience of the reader. Note also that there is commented out code in the progams 
  for these other formats.)

RATS PROCEDURE FILES: bkfilter-1.src, bndecomp.src (RATS procedure files)
* The file bkfilter-1.src is needed to run the program bk6614.prg given above. 
* The file bndecomp.src is needed to run the program bn6614.prg given above.
* These two files are RATS procedure files that were downloaded from the site:
  www.estima.com/cgi-bin/procbrowser.cgi?Version=900&Revision=Any&Subject=Any&Reference=Any

OUTPUT FILES: hp6614_all.db, linquddum_all.db, wat6614_all.db, hc6614_all.db, 
                             bk6614_all.db, bn6614_all.db (fame format)
              hp6614_all.rat, linquddum6614_all.rat, wat6614_all.rat, hc6614_all.rat, 
                              bk6614_all.rat, bn6614_all.rat (rats format)
              hp6614_all.xls, lin6614_all.xls, qud6614_all.xls, dum6614_all.xls, 
              wat6614_all.xls, hc6614_all.xls, bk6614_all.xls,  bn6614_all.xls 
                                                                (excel format)
* The hp6614_all.___ output files are the output of the program file hp6614.prg and all 
  files are identical to each other; just saved in different formats. Each file includes 
  198 series, all but one of the series have names of the form "HP_GAP_GDPXXXX" The series 
  that does not have this form is "HP_GAP_RT" which provides the real-time estimates of the
  HP filter generated output gap. For all the other series in the file the "XXXX" in series 
  name "HP_GAP_GDPXXXX" is a date; specifically, any quarter between "96Q1" and 14Q1".
  These series are full timeseries of the output gap generated by the HP filter using the
  GDPXXXX vintage of real GDP data described above.
* The wat6614_all.___ output files are the output of the program file wat6614.prg and all 
  files are identical to each other. These output files have the same number of series as
  hp6614_all.___ and the series are similarly named. The difference in these files is that 
  all series have the prefix "WAT".
* The hc6614_all.___ output files are the output of the program file hc6614.prg and all 
  files are identical to each other. These output files have the same number of series as
  hp6614_all.___ and the series are similarly named. The difference in these files is that 
  all series have the prefix "HC".
* The bk6614_all.___ output files are the output of the program file bk6614.prg and all 
  files are identical to each other. These output files have the same number of series as
  hp6614_all.___ and the series are similarly named. The difference in these files is that 
  all series have the prefix "BK".
* The bn6614_all.___ output files are the output of the program file bn6614.prg and all 
  files are identical to each other. These output files have the same number of series as
  hp6614_all.___ and the series are similarly named. The difference in these files is that 
  all series have the prefix "BN".
* The linquddum6614_all.db, linquddum6614_all.rat, lin6614_all.xls, qud6614_all.xls, and 
  dum6614_all.xls output files are the output of the program file lqd6614_all.prg.  The 
  files lqd6614_all.db and lqd6614_all.rat are identical to each other. Because excel files 
  cannot include as many variable as fame and rats databases the series included in 
  linquddum6614_all.db and linquddum6614_all.rat are split across three excel databases 
  lin6614_all.xls (for the linear trend), qud6614_all.xls (for the quadratic trend), and 
  dum6614_all.xls (for the broken trend).  These output files have the same number of series
  given the detrending method and the series are similarly named. The difference in these 
  files is that all linear detrended series have the prefix "LIN", all quadratic detrended 
  series have the prefix "QUD", and all broken trend detrended series have the prefix 
  "DUM".

********************************************************************************************

SECOND STEP IN GENERATING THE RESULTS IN TABLE 2

DATA FILES: hp6614_all.db, linquddum_all.db, wat6614_all.db, hc6614_all.db, 
                           bk6614_all.db, bn6614_all.db (fame format)
            hp6614_all.rat, linquddum6614_all.rat, wat6614_all.rat, hc6614_all.rat, 
                            bk6614_all.rat, bn6614_all.rat (rats format)
            hp6614_all.xls, lin6614_all.xls, qud6614_all.xls, dum6614_all.xls, 
            wat6614_all.xls, hc6614_all.xls, bk6614_all.xls,  bn6614_all.xls 
                                                              (excel format)
* These files are the "OUTPUT FILES" for the "FIRST STEP IN GENERATING THE RESULTS IN 
  TABLE 2" (listed above).

PROGRAM FILES: table2stats.inp (fame program)
* This program calculates the noise-to-signal ratios shown in Table 2 for the Hodrick-
  Prescott filter, Broken trend, Quadratic trend, Linear trend, Watson Kalman filter model, 
  Harvey and Clark Kalman filter model, Baxter King band-pass filter model, and Beveridge-
  Nelson trend-cycle decomposition and does so over a number of different sample periods.
* The program uses the data files hp6614.db, lqd6614.db, wat6614.db, hc6614.db, bk6614.db, 
  bn6614.db for its calculations. (It does not use the files in the other formats but these 
  are provided for the convenience of the reader.) 

OUTPUT FILES: table2stats_output.txt (text file)
* This file provides the output from table2stats.inp. The numbers generated by 
  table2stats.inp and shown in table2stats_output.txt correspond to those in Table 2.


********************************************************************************************
********************************************************************************************
********************************************************************************************


DATA FILES AND PROGRAMS FOR PERFORMING THE STATISTICAL SIGNIFICANCE TESTS ON THE RESULTS 
IN TABLES 1 AND 2 USING THE "SECOND" (REGRESSION-BASED) APPROACH DISCUSSED IN SECTION III
AND FOOTNOTE 15 OF THE PAPER

There are three steps involved in performing the statistical significance tests on the 
results in table 1 and 2 using the "second" (regression-based) approach used in Section III
and footnote 15. The first two steps involve constructing the data that are used for the
tests; the third step involves performing the regression-based tests themselves.

********************************************************************************************

FIRST STEP, WHICH INVOLVES MAKING THE DATA SERIES -- CALLED X-TILDE AND Z-TILDE -- USED IN 
REGRESSION (1)

DATA FILES: gbgap_all.db, hp6614_all.db, linquddum_all.db, wat6614_all.db, 
                          hc6614_all.db, bk6614_all.db, bn6614_all.db (fame format),
            gbgap_all.rat, hp6614_all.rat, linquddum6614_all.rat, wat6614_all.rat, 
                           hc6614_all.rat, bk6614_all.rat, bn6614_all.rat (rats format),
            gbgap_all.xls, hp6614_all.xls, lin6614_all.xls, qud6614_all.xls, 
                           dum6614_all.xls, wat6614_all.xls, hc6614_all.xls, 
                           bk6614_all.xls,  bn6614_all.xls (excel format)
* These data files are the "DATA FILES" listed above for "PROGRAMS AND DATA FILES FOR 
  GENERATING THE RESULTS IN TABLE 1" and the "OUTPUT FILES" listed above "FIRST STEP IN 
  GENERATING THE RESULTS IN TABLE 2"  
* These data files and the variables they include are described above.

PROGRAM FILES: signoise_13yrs.prg, signoise_32yrs.prg, signoise_9yrs.prg 
               (all rats programs)
* Each of the program files with the names signoise___.prg calculate the x-tilde and 
  z-tilde series used in regression (1).
* The program signoise_13yrs.prg calculates the x-tilde and z-tilde series for the variant 
  of regression (1) that tests the statistical significance of the differences between the
  signal-to-noise ratios calculated over 1980-1992 and over 1994-2006.  The same program 
  calculates the observations for x-tilde and z-tilde over both parts of the sample (that 
  is, over both 1980-1992 and 1994-2006).
* The programs signoise_32yrs.prg and signoise_9yrs.prg calculates the x-tilde and z-tilde 
  series for the variant of regression (1) that tests the statistical significance of 
  differences between the signal-to-noise ratios calculated over 1966-1997 and over 
  1998-2006. The program signoise_32yrs.prg calculates the observations for x-tilde and
  z-tilde over the earlier part of the sample (that is, over 1966-1997) and program 
  signoise_9yrs.prg calculates the observations for x-tilde and z-tilde over the later
  part of the sample (that is, over 1998-2006).
* The programs use the data files gbgap_all.rat, hp6614_all.rat, linquddum6614_all.rat, 
  wat6614_all.rat, hc6614_all.rat, bk6614_all.rat, bn6614_all.rat for their calculations. 
  (The programs do not use the files in the other formats, which are provided for the 
  convenience of the reader. Note also that there is commented out code in the progams 
  for these other formats.)

OUTPUT FILES: revgapsq_fnv_13yrs.db, revgapvr_fnv_13yrs.db, 
              revgapsq_fnv_32yrs.db, revgapvr_fnv_32yrs.db, 
              revgapsq_fnv_9yrs.db, revgapvr_fnv_9yrs.db (fame format),   
              revgapsq_fnv_13yrs.rat, revgapvr_fnv_13yrs.rat, 
              revgapsq_fnv_32yrs.rat, revgapvr_fnv_32yrs.rat, 
              revgapsq_fnv_9yrs.rat, revgapvr_fnv_9yrs.rat (rats format),  
              revgapsq_fnv_13yrs.xls, revgapvr_fnv_13yrs.xls, 
              revgapsq_fnv_32yrs.xls, revgapvr_fnv_32yrs.xls, 
              revgapsq_fnv_9yrs.xls, revgapvr_fnv_9yrs.xls (excel format)  
* The revgapsq_fnv_13yrs.___ and revgapvr_fnv_13yrs.___ output files are the output of the 
  program file signoise_13yrs.prg and all revgapsq_fnv_13yrs.___ files are identical to 
  each other and all revgapvr_fnv_13yrs.___ are identical to each other; just saved in 
  different formats. Each revgapsq_fnv_13yrs.___ file includes 9 series, all of which have 
  names of the form "XX_REVGAPSQ_F13". These variables are the x-tilde series described in 
  Section III of the paper series. Similarly each revgapsq_fnv_13yrs.___ file includes 9 
  series, all of which have names of the form "XX_REVGAPVR_F13" These variables are the 
  z-tilde series described in Section III of the paper series. (Note that in the text in 
  the paper the description is just for the Greenbook output gap, but the description 
  applies to all other output gap measures too). The "XX" in the series names 
  "XX_REVGAPSQ_F13" and "XX_REVGAPVR_F13" correspond to the method used to obtain the 
  output gap measure for which the x-tilde or z-tilde series is constructed and, in 
  particular, use the following code: gb = Greenbook, hp = Hodrick-Prescott filter, lin =
  linear detrending method, qud = quadratic detrending method, dum = broken trend 
  detrending method, wat = Watson Kalman filter model, hc = Harvey and Clark Kalman filter 
  model, bk = Baxter King band-pass filter model, and bn = Beveridge-Nelson trend-cycle 
  decomposition.
* The revgapsq_fnv_32yrs.___ and revgapvr_fnv_32yrs.___ output files are the output of the 
  program file signoise_32yrs.prg and all revgapsq_fnv_32yrs.___ files are identical to 
  each other and all revgapvr_fnv_32yrs.___ are identical to each other; just saved in 
  different formats. These output files have the same number of series as 
  revgapsq_fnv_13yrs.___ and revgapvr_fnv_13yrs.___ and the series are similarly named. 
  The difference in these files is that the series all have the suffix "F32".
* The revgapsq_fnv_9yrs.___ and revgapvr_fnv_9yrs.___ output files are the output of the 
  program file signoise_9yrs.prg and all revgapsq_fnv_9yrs.___ files are identical to 
  each other and all revgapvr_fnv_9yrs.___ are identical to each other; just saved in 
  different formats. These output files have the same number of series as 
  revgapsq_fnv_13yrs.___ and revgapvr_fnv_13yrs.___ and the series are similarly named. 
  The difference in these files is that the series all have the suffix "F09".

********************************************************************************************

SECOND STEP, WHICH INVOLVES (A) CONSOLIDATING THE OUTPUT FILES GENERATED BY THE PROGRAMS
signoise_13yrs.prg, signoise_32yrs.prg, AND signoise_9yrs.prg INTO JUST THE X-TILDE AND
Z-TILDE SERIES FOR WHICH STATISTICAL SIGNIFICANCE EXAMINED AND (B) MAKING SEPARATE OUTPUT
FILES FOR TABLE 1 SIGNIFICANCE TESTS AND TABLE 2 SIGNIFICANCE TESTS

DATA FILES: revgapsq_fnv_13yrs.db, revgapvr_fnv_13yrs.db, 
            revgapsq_fnv_32yrs.db, revgapvr_fnv_32yrs.db, 
            revgapsq_fnv_9yrs.db, revgapvr_fnv_9yrs.db (fame format),   
            revgapsq_fnv_13yrs.rat, revgapvr_fnv_13yrs.rat, 
            revgapsq_fnv_32yrs.rat, revgapvr_fnv_32yrs.rat, 
            revgapsq_fnv_9yrs.rat, revgapvr_fnv_9yrs.rat (rats format),  
            revgapsq_fnv_13yrs.xls, revgapvr_fnv_13yrs.xls, 
            revgapsq_fnv_32yrs.xls, revgapvr_fnv_32yrs.xls, 
            revgapsq_fnv_9yrs.xls, revgapvr_fnv_9yrs.xls (excel format) 
* These are the "OUTPUT FILES" listed above under the heading of the "FIRST STEP, WHICH 
  INVOLVES MAKING THE DATA SERIES USED IN REGRESSION (1) ..."

PROGRAM FILES: mk_gb_reg_data.prg and mk_stat_reg_data.prg (all rats programs)
* The programs use the data files revgapsq_fnv_13yrs.rat, revgapvr_fnv_13yrs.rat, 
  revgapsq_fnv_32yrs.rat, revgapvr_fnv_32yrs.rat, revgapsq_fnv_9yrs.rat, and 
  revgapvr_fnv_9yrs.rat. (The programs do not use the files in the other formats, which 
  are provided for the convenience of the reader. Note also that there is commented out 
  code in the programs for these other formats.)
* The program "mk_gb_reg_data.prg" reads in the data files revgapsq_fnv_13yrs.rat, 
  revgapvr_fnv_13yrs.rat, revgapsq_fnv_32yrs.rat, revgapvr_fnv_32yrs.rat, 
  revgapsq_fnv_9yrs.rat, and revgapvr_fnv_9yrs.rat and saves all Greenbook related series 
  -- specifically, series that begin with the letters "gb" -- to a database for testing
  Table 1 significance.
* The program "mk_stat_reg_data.prg" reads in the data files revgapsq_fnv_13yrs.rat, 
  revgapvr_fnv_13yrs.rat, revgapsq_fnv_32yrs.rat, revgapvr_fnv_32yrs.rat, 
  revgapsq_fnv_9yrs.rat, and revgapvr_fnv_9yrs.rat and saves a subset series to a database 
  for testing Table 2 significance.

OUTPUT FILES: table1_reg_test_data.db, table2_reg_test_data.db (fame format),
              table1_reg_test_data.rat, table2_reg_test_data.rat (rats format),
              table1_reg_test_data.xls, table2_reg_test_data.xls (excel format)
* The table1_reg_test_data.___ output files are the output of the program file 
  mk_gb_reg_data.prg and all files are identical to each other; just saved in different 
  formats. Each file includes the following six series: GB_REVGAPSQ_F13, GB_REVGAPVR_F13,
  GB_REVGAPSQ_F32, GB_REVGAPVR_F32, GB_REVGAPSQ_F09, and GB_REVGAPVR_F09. As noted above
  series with the letters "SQ" in their names correspond to x-tilde variables while series 
  with "VR" in their names correspond to z-tilde variables.
* The table2_reg_test_data.___ output files are the output of the program file 
  mk_gb_reg_data.prg and all files are identical to each other; just saved in different 
  formats. For the statistical detrending methods we only perform the statistical 
  significance analysis for the output gap measures that show a declines in both their 
  standard-deviation based and RMSE-based signal-to-noise ratios between the earlier and 
  later sample periods.  As a result file only includes the series BK_REVGAPSQ_F13, 
  BK_REVGAPVR_F13, HP_REVGAPSQ_F32, HP_REVGAPVR_F32, HC_REVGAPSQ_F32, HC_REVGAPVR_F32,
  BK_REVGAPSQ_F32, BK_REVGAPVR_F32, BN_REVGAPSQ_F32, BN_REVGAPVR_F32, HP_REVGAPSQ_F09,	
  HP_REVGAPVR_F09, HC_REVGAPSQ_F09, HC_REVGAPVR_F09, BK_REVGAPSQ_F09, BK_REVGAPVR_F09,
  BN_REVGAPSQ_F09, and BN_REVGAPVR_F09.

********************************************************************************************

THIRD STEP, WHICH INVOLVES RUNNING THE DUMMY-VARIABLE REGRESSIONS

PROGRAM FILES FOR TESTING THE TABLE 1 RESULTS: There are two sets of RATS programs that run
the dummy-variable regressions (regression 1) that we use to assess the statistical
significance of the improvements in NSRs that we observe in Table 1.  The first set (which
consists of a single program) uses HAC-corrected standard errors to compute the relevant
t-statistic, while the second set (which consists of four programs) implements the naive
block bootstrap described in footnote 11.
* The program mean_tests1.prg uses the dataset table1_reg_test_data.rat from step 2, above,
  and runs all of the dummy variable regressions.
* The following four programs use the dataset table1_reg_test_data.rat from step 2, above,
  to implement the bootstrap routine:
  **  rmse_boot_80_92v94_06.prg:   RMSE-based NSRs, 1980-92 sample vs. 1994-2006 sample.
  **  stddev_boot_80_92v94_06.prg: Standard-deviation-based NSRs, 1980-92 sample vs.
                                   1994-2006 sample.
  **  rmse_boot_66_97v98_06.prg:   RMSE-based NSRs, 1966-97 sample vs. 1998-2006 sample.
  **  stddev_boot_66_97v98_06.prg: Standard-deviation-based NSRs, 1966-97 sample vs.
                                   1998-2006 sample.

OUTPUT FILES: Files with the extension .out contain the output from the corresponding
program file.  Specifically, mean_tests1.out contains the regression output from program
mean_tests1.prg, while rmse_boot_80_92v94_06.out, stddev_boot_80_92v94_06.out, etc. contain
the output from programs rmse_boot_80_92v94_06.prg, stddev_boot_80_92v94_06.prg, etc.
Note that the output from this latter set of programs consists of the dummy variable
regression output (which will match the results for the corresponding regression from
mean_tests1.out), along with the tabulated bootstrapped t-statistics (which can be compared
to the t-statistic for the dummy variable to assess its statistical significance).


PROGRAM FILES FOR TESTING THE TABLE 2 RESULTS: The RATS programs that run the
dummy-variable regressions (regression 1) that we use to assess the statistical
significance of the NSRs in Table 2 from the statistical detrending procedures that
manifest consistent improvement have filenames that take the form
XX_YYYY_signif_PERIOD.prg, where:

XX:     is the detrending method used (HP = Hoderick-Prescott, HC = Harvey-Clark,
        BK = Baxter-King, BN = Beveridge-Nelson);
YYYY:   denotes which definition of the NSR is considered (RMSE = RMSE-based NSR,
        STDDEV = standard-deviation-based NSR); and
PERIOD: gives the sample period over which the improvement is being assessed
        (6697_9806 = 1966-97 vs. 1998-2006, 8092_9406 = 1980-92 vs. 1998-2006).

For example, program bk_rmse_signif_6697_9806.prg assesses the improvement in the
RMSE-based definition of the NSR for the Baxter-King-based output gap measure between
the 1966-97 subperiod and the 1998-2006 subperiod.

Note that these programs all use the dataset table2_reg_test_data.rat from step 2, above;
in addition, each program runs both the dummy variable regression (with HAC-corrected
standard errors) and the corresponding naive block bootstrap routine.

OUTPUT FILES: Files with the extension .out contain the output from the corresponding
program file.  (Each program's output file includes the results from the relevant dummy
variable regression along with the tabulated bootstrapped t-statistics.)


********************************************************************************************
********************************************************************************************
********************************************************************************************


DATA FILE AND PROGRAM FOR GENERATING THE RESULTS IN TABLE 3

DATA FILE: table3_data.rat (RATS format)
* table3_data.rat is a RATS database that contains the following series:
  GAPMMMYY:     Time series of the output gap from the month mmm, year yy Greenbook
  P_PCEX:       Core PCE chain price index
  P_COREIMP:    Core nonfuel imports chain price index
  NOM_COREIMP:  Nominal core nonfuel imports
  NOM_COREPCE:  Nominal core PCE
  The output gap series are obtained from Fed databases.  The core PCE price index and
  nominal core PCE series were downloaded from the Bureau of Economic Analysis (BEA)
  website on April 27, 2012.  The price index for core nonfuel imports is constructed by
  Fed staff using a Fisher aggregation routine that replicates the procedure followed by
  the BEA in constructing the national accounts; specifically, the measure is defined by
  stripping out the price indexes for imports of natural gas, computers, and
  semiconductors from the published price index for imports of nonoil goods.  Likewise,
  nominal core nonfuel imports are defined as nominal nonoil goods imports less nominal
  imports of natural gas, computers, and semiconductors.  (The source for all of these
  series is the BEA.)

PROGRAM FILE: table3_repl1.prg (RATS program)
* This program uses the data from table3_data.rat to estimate the various forecasting
  models and dynamic simulations (all data transformations are performed by the program
  code), and uses the results from the simulations to calculate the RMSEs shown in
  Table 3.

OUTPUT FILE: table3_repl1.out (text file)
* This file provides the output from table3_repl1.prg (specifically, the RMSEs shown in
  Table 3).

********************************************************************************************


[END OF README FILE]
